OpenVR Benchmark Shows Unknown AMD Radeon GPU Beating RTX 2080 Ti by 17 Percent
AMD had a great start to 2020, but one product gamers had hoped to see — so-called “Big” Navi — was only briefly mentioned by CEO Dr. Lisa Su. We know that Big Navi is absolutely in the works, but not much more than that. Here’s Lisa Su on the topic:
I know those on Reddit want a high end Navi! You should expect that we will have a high-end Navi, and that it is important to have it. The discrete graphics market, especially at the high end, is very important to us. So you should expect that we will have a high-end Navi, although I don’t usually comment on unannounced products.
Now we’ve got benchmarks tipping up showing an unknown Radeon beating the RTX 2080 Ti in OpenVR’s benchmark at a resolution of 1512×1680. AMD’s unknown Radeon comes in second place, with a score of 103.32, while the fastest RTX 2080 Ti checks in at 88.1. Result? AMD wins by roughly 17 percent.
What does this fundamentally mean for AMD’s upcoming “Big Navi?” Nothing. We don’t know how optimized the drivers are, or whether the unknown Radeon actually rendered all of the scenes correctly. We don’t know the clock or memory configuration on the GPU. Recent rumors have suggested that Big Navi is twice the size of Navi, with up to 80 CUs and 5120 cores. The problem with these kinds of leaks, however, is that they only represent performance in a single moment of time (and with a GPU that may not be anywhere near production-ready).
But there’s a separate issue here, common to any situation where one benchmark is used to claim dominance. In the slideshow below, I’ve clipped off most of the graphs and focused solely on the RTX 2070 versus the 5700 XT. Watch how the performance narrative changes depending on which game I test:
I picked these graphs because they demonstrate how much comparative performance can change from title to title. Winning the OpenVR benchmark by 17 percent over the 2080 Ti is good, inasmuch as it demonstrates AMD is building a competitive GPU. But — as the slideshow above illustrates — GPUs can be fabulous in some workloads and struggle with others. RDNA has proven to have chops as a rendering and compute GPU and I expect Big Navi will deliver in that regard.
If I had to guess, based on the information publicly available, I’d guess that Big Navi will be solidly faster than Turing. I don’t know how it will compare against Ampere. Nvidia has historically had good luck with node shifts. There are claims that Ampere could be up to 50 percent faster than Turing. Those claims should be taken with a large grain of salt, but we do expect Big Navi to face real competition from Nvidia’s first 7nm GPUs in 2020.